RAPM VORP Spreadsheet & Analysis

[LA Bird]

+1 for all the work that went into this but I'll share some of my thoughts.

Methodology:
- The pre PBP projection could be improved with a BPM adjustment or something similar after adjusting for variance. For a basic example, Mark Price having a 4.3 BPM before 1997 vs 0.5 after means we can add the +3.8 difference to his post 1997 RAPM for a bit more accurate career projection. Obviously that means the number is no longer purely RAPM but I think it's still more meaningful than an extrapolation on way past prime seasons.
- Unlike with box scores, a X year career RAPM VORP is not necessarily the same as the summation of n distinct X/n year RAPM VORPs. It's good to have such a large career dataset for an all-in-one metric but I personally find the individual sliding windows to be more insightful. Any of the 3~5 year models are fine IMO and it's up to how you want to balance between having enough years for the data to stabilize while at the same time not having so many that the earlier seasons in the sample get drowned out.
- Going against what is long established consensus here but measuring player value against replacement linearly doesn't really make sense to me. For example, there is no way you can convince me 1 season of +10 LeBron (+13.3 above -3.3 replacement) is equal to 3 seasons of +1.1 DeRozan (3x4.4=13.2). If we look at the bell curve for BPM:



The relative frequency of a +1 season and a +10 season is not remotely near 3 to 1. It's like 3 standard deviations to 1 if not more. Comparing against a negative baseline only further devalues the top seasons disproportionately.


Player results:
- Chris Paul and Paul Pierce have some of the largest RS to PO RAPM drop-offs so they can get overrated in a RS only dataset like this, especially in a cumulative stat where their longevity advantage amplifies per possession numbers. CP3 is still quite elite in playoffs RAPM despite the dropoff but, FWIW, Pierce falls to almost a zero on a large playoffs sample.
- Besides these RAPM numbers, Divac is 30th in career WOWYR/GPM average, 95+ percentile in RWOWY on both the Lakers and Kings, and he is 2nd in raw on/off in 95 and partial RAPM in 91. Impact wise, Divac has a consistently strong record. And while I feel like his O/DRAPM split leans too much towards defense, we are still talking about a fairly mobile 7 footer who peaked at 1.6 steals and 2.2 blocks. Add in his case as arguably the GOAT passing center besides Jokic and it's a surprise he is not top 100 more often. Unseld is a better rebounder but otherwise, Divac has a similar case. Also random trivia: Divac is the only player to outsteal and outblock prime Hakeem in a playoff series.
- Conley is one of the best players relative to accolades but PG is a tough position to crack for someone without a USP. Baron Davis probably has a stronger case with some notable playoffs victories but Gus Williams barely made the top 100 even with a ring and his argument is similar when it comes to longevity and playoffs elevation. Personally I would prefer both over Tony Parker last project but 4 rings is tough to beat.
- Billups' career impact stats is low key pretty mediocre outside of the 06-10 stretch. He was never a big defensive stopper and pre-Saunders, he doesn't appear to be having that much of an offensive impact either. On a career scale, I wonder if Billups gets overrated a little on this board because he is the ideal complementary PG fit next to superstar wings (at least before Curry came around).
- Iguodala is an interesting case to me because he had a very notable transition from star to role player. If we look at the 3 year RAPMs on nbarapm.com, he mostly hovered around +1 in his Philly years. It wasn't until he shot less and focused on defense that his RAPM skyrocketed. Multiplying his career minutes by his career RAPM would thus be a misrepresentation of his true career VORP given that he was largely ineffective as the lead star for years.

[Doctor MJ]

Why does Hakeem have a -2 average offensive RAPM? Sorry if you already mentioned this, but what year does it start?

It starts with '96-97. which also happens to be the year Charles Barkley arrived in Houston. I believe that basically what happened is that between him losing a step and adding a teammate who maybe always had a superior regular season offensive attack, his continued primacy in scoring actually held the superteam back.

Basically, imho, what you see there basically has pretty small relevance to estimating the impact of Hakeem in the playoffs in those title runs...but it does also give us into a window of Olajuwon's offensive impact being less resilient to similar perturbations than, say, Barkley's. Which is of course fascinating because we saw how exceptionally resilient his offense seemed against the best to get the chips. Seems like a contradiction, but I'd argue it isn't. There are different dimensions to resilience.

This also to say that I do think Hakeem would look a lot better if we had all his regular season data, but it's really the playoff data we'd need to study most intently.

[Doctor MJ]

Where is Jeff Foster?

He's in there. Let's see: 191st, between Dale Davis & Mike Miller.

[Top10alltime]

Cool

[Doctor MJ]

+1 for all the work that went into this but I'll share some of my thoughts.

Methodology:
- The pre PBP projection could be improved with a BPM adjustment or something similar after adjusting for variance. For a basic example, Mark Price having a 4.3 BPM before 1997 vs 0.5 after means we can add the +3.8 difference to his post 1997 RAPM for a bit more accurate career projection. Obviously that means the number is no longer purely RAPM but I think it's still more meaningful than an extrapolation on way past prime seasons.
- Unlike with box scores, a X year career RAPM VORP is not necessarily the same as the summation of n distinct X/n year RAPM VORPs. It's good to have such a large career dataset for an all-in-one metric but I personally find the individual sliding windows to be more insightful. Any of the 3~5 year models are fine IMO and it's up to how you want to balance between having enough years for the data to stabilize while at the same time not having so many that the earlier seasons in the sample get drowned out.
- Going against what is long established consensus here but measuring player value against replacement linearly doesn't really make sense to me. For example, there is no way you can convince me 1 season of +10 LeBron (+13.3 above -3.3 replacement) is equal to 3 seasons of +1.1 DeRozan (3x4.4=13.2). If we look at the bell curve for BPM:



The relative frequency of a +1 season and a +10 season is not remotely near 3 to 1. It's like 3 standard deviations to 1 if not more. Comparing against a negative baseline only further devalues the top seasons disproportionately.


Player results:
- Chris Paul and Paul Pierce have some of the largest RS to PO RAPM drop-offs so they can get overrated in a RS only dataset like this, especially in a cumulative stat where their longevity advantage amplifies per possession numbers. CP3 is still quite elite in playoffs RAPM despite the dropoff but, FWIW, Pierce falls to almost a zero on a large playoffs sample.
- Besides these RAPM numbers, Divac is 30th in career WOWYR/GPM average, 95+ percentile in RWOWY on both the Lakers and Kings, and he is 2nd in raw on/off in 95 and partial RAPM in 91. Impact wise, Divac has a consistently strong record. And while I feel like his O/DRAPM split leans too much towards defense, we are still talking about a fairly mobile 7 footer who peaked at 1.6 steals and 2.2 blocks. Add in his case as arguably the GOAT passing center besides Jokic and it's a surprise he is not top 100 more often. Unseld is a better rebounder but otherwise, Divac has a similar case. Also random trivia: Divac is the only player to outsteal and outblock prime Hakeem in a playoff series.
- Conley is one of the best players relative to accolades but PG is a tough position to crack for someone without a USP. Baron Davis probably has a stronger case with some notable playoffs victories but Gus Williams barely made the top 100 even with a ring and his argument is similar when it comes to longevity and playoffs elevation. Personally I would prefer both over Tony Parker last project but 4 rings is tough to beat.
- Billups' career impact stats is low key pretty mediocre outside of the 06-10 stretch. He was never a big defensive stopper and pre-Saunders, he doesn't appear to be having that much of an offensive impact either. On a career scale, I wonder if Billups gets overrated a little on this board because he is the ideal complementary PG fit next to superstar wings (at least before Curry came around).
- Iguodala is an interesting case to me because he had a very notable transition from star to role player. If we look at the 3 year RAPMs on nbarapm.com, he mostly hovered around +1 in his Philly years. It wasn't until he shot less and focused on defense that his RAPM skyrocketed. Multiplying his career minutes by his career RAPM would thus be a misrepresentation of his true career VORP given that he was largely ineffective as the lead star for years.

Really appreciate the thoughts LA Bird!

- You're right that I could use BPM to more accurately project into the past.

- Problematic use of career average rather than focusing on shorter epochs that represent prime/peak. Oh definitely, and the next project I expect to do will be focused on bin-ing such epochs. That will serve as a nice contrast to the career average approach.

- 1 LeBron year != 3 DeRozan years. I certainly agree, but it's a question of "Why?" when considering adjustment. One might say "it's like 9 years rather than 3" which could be achieved with different replacement levels, but I think that we both see a more general issue:

The NBA's entire culture is built around not just being proficient at basketball, but being the best in the world over a tournament of series in which opponents have ample time to strategize against star players, and while this arguably results in LeBron becoming an even greater outlier with DeRozan it's fool's gold.

DeRozan thus isn't actually giving a franchise simply a fractional positive impact that taken across a long enough duration could lead to more net value than a better player over a shorter period of time. He's setting a ceiling on your team to below contender levels. He's a great attractor toward the proverbial treadmill. So, how many years of treadmill will it take before DeRozan's career amounts to more than LeBron's best single year? False premise in my assessment. No amount of treadmill is more valuable a single chip.

Others can disagree of course and I think they should speak their mind, but if you and I are in agreement there, then I think it's worth noting we might mostly agree on the big things.

Soon enough I intend to get into slices such as Draft, and DeRozan's place in the 2009 draft. I don't want to go into too much detail here without having prepared a good table, but while DeRozan's all-star stature would put him 4th in the draft (Curry, Harden, Griffin) - and his longevity can lead to arguments for him to rank 3rd (above Griffin) - I've always had DeRozan way lower than that.

So the fact that when you see the 2009 draft cohort DeRozan ranks 4th (Curry, Harden, Jrue), know that that raised my eyebrows. It's something my replacement level definitely affects (more negative replacement level effectively weighs longevity more strongly), so perhaps an argument for a different level...but the thing is, that treadmill factor will always be there, and in the end, that's a reason not to use metrics in the final pass of your analysis.

- Paul; Pierce; Playoffs. Definitely worth doing something like this for playoff data (as stated, I'd like to see a sample that includes 2025). I do think that that data is always going to be pretty dicey because sample size is so small (max 1 opponent per round), but I'd like to see the data laid out consider it myself.

- Divac great! Indeed, preach!

- Conley, Baron, Parker. So I'd say I'm roughly with you. I'm not really expecting to champion Conley or Baron as must-100 guys...but they do look better than Tony Parker, and I don't think Parker really has a clear cut case as being better in the playoffs despite the chips. None of this means Parker wasn't a good, valuable player, because he certainly was, but him being "tiered" well ahead of Conley & Baron isn't something I see as analytically justified.

- Billups. I think he's someone who is definitely benefitting from the way we try to latch onto players to represent team success. Before the 2004 NBA Finals, most would have said that likely no Piston on that team would make the Hall. Afterward, while there was certainly debate, I saw Ben as a likely HOFer, but did not think anyone else was likely. But Billups winning the Finals MVP and then emerging as the star of a team that continued contending post-Ben, really changed things. Then the noteworthy success in Denver sure seemed to solidify him as a lock.

None of this means Billups is necessarily overrated, so much as that there's more going into how people are weighing Billups career than I would expect to show in data like this.

- Iggy. The idea of a guy's career RAPM being misleading because he was super-valuable in one role but not another, is certainly true. Iggy should not be your offensive alpha.

I think we need to be clear though that Iggy was scoring 18 PPG with Iverson scoring 30+, and only went up to 19.9 PPG post-Iverson. So while he did lead the team in scoring post-Iverson, he was still basically playing with the same (secondary) primacy.

Now, none of this changes the truth you point out about Iggy being most valuable after he reduces his scoring primacy down a lot further than that, but I do think it makes sense to push back when you say 'largely ineffective as the lead star for years'. The reality is that when Iggy was seen as the star of his team, the team's offensive strategy was essentially star-less distributing the volume across teammates.

Thus, I'd argue that trying to organize Iggy's career into something like RP/Star/RP with him clearly going from good to bad and back to good depending on his scoring focus is over-simplistic. I think if we just look at the rosters on the 76ers post-AI, it's not like they had the Splash Brothers there just waiting in the wings. This then to say, that I don't think Iggy scoring way less on those 76ers team would have helped them. I think he only works best for a team on super-low scoring primacy if his teammates are just that good.

Going along with this, I'd emphasize that there's more to Iggy's defensive run in GS than just him being more defensively focused in his role - meaning, it's not like I think Iggy was capable of GS levels of defensive impact on any NBA team as long as he didn't have to score double digits. I think what we see in GS is very much about something in that context, and it's up to us what we think about it.

[Smoothbutta]

Simpleton question here doing a sniff test but how is someone like Conley or Sheed or Vlade ahead of Giannis or Kawhi?

Also is it possible to have a different breakdown for 3-year stint peak or so? Or it would be nice to be able to limit the data so it only looks at ages 23-34 for example rather than including years that bring a player down like age 37 KG or wizard's MJ etc.

[Doctor MJ]

Update:

I've added a second TweakMe set for an alternative replacement value (Alt RP)

So I now have two tabs "RP Raw" and "Alt RP Raw" that are the same thing, but using 2 separate RP levels.

I've also update the "RealGM 100 Perspective" tab specifically so that we can see the calculations for VORP based on both sets.

So then, when we go from my current RP (-3.3), to something that I would generally say is on the other end of common RP spectrum (-2.0), changes we see on this tab:

- Numbers are lowered generally, and so, for example, the #1 LeBron goes from 7856 to 7088.

- I've used 1000 as a recognizable standard for a notable level of accumulated value over a career. Using the current RP level, I've found 176 player careers (with projection) that break that threshold. With the Alt RP level, this number down to 121. (Not a good or bad thing, just a thing.)

If you look at the lists side by side on the RealGM 100 Perspective tab, just expect that with the less forgiving Alt RP, all numbers will be smaller, but that players whose competitive edge was based on longevity will get dropped harder, and so you can have shorter-playing guys with higher RAPM "rise".

Alright, I'll add one other query that's important to me, and that's the guys who had long careers as apparent below Replacement Level guys. Basically, I just think it's a big deal if these guys are negative, and so having two differing lists based on RP level is useful to me:

So, among guys with at least 20K non-projected MP:

For -3.3 RP level, the negative, from most to least:

Jeff Green
Ricky Davis
Drew Gooden
Tristan Thompson

For the -2.0 Alt RP level:

Jeff Green
Ricky Davis
Drew Gooden
Juwan Howard
Jarrett Jack
Tristan Thompson
Luke Ridnour
Jason Williams
Samuel Dalembert
John Salmons
Larry Hughes
DJ Augustin

So, just to keep in mind, in the PBP Era, we've had 219 guys so far break the 20K RS MP threshold.

The -3.3 RP level classifies 4 of these 219 as below replacement level.
The -2.0 RP level classified 12.

I think it's absolutely worth discussing whether we feel like these guys were below replacement level players, and why it was they played long careers at such problematic levels. (By contrast, I take it as a given that plenty of would-be stars flame out with below RP level careers. It only gets particularly interesting when we think the data says that's what should have happened for a guy...but somehow he got to play 20K mins, which required coaches continuing to choose to play him for quite a long time.)

I'll pause before going any further with stuff.

I'll say it's my intent to keep 2 RP levels handy from now on so we have a sense for much our uncertainty in RP is having an effect on our perception when we choose significantly differing levels - but I'm really not set on what those levels are, only that I'd like to make sure we capture the differing perspectives of people who care about this (we sickos).

[Doctor MJ]

Simpleton question here doing a sniff test but how is someone like Conley or Sheed or Vlade ahead of Giannis or Kawhi?

Also is it possible to have a different breakdown for 3-year stint peak or so?

So first, just know that we've got some really nice sources for things like 3-year studies now - relatively suddenly in the last year or so. While I intend to do and share similar analyses based on them, please do go to the sites:

https://www.nbarapm.com/
https://www.thebasketballdatabase.com/

Keep in mind that there's not one universal standard for this data so you should expect slight differences between sources, as well as more noise in something like a 3-year study compared to lifetime.

Re: How Conley, Sheed, Vlade ahead of Giannis or Kawhi?

The short answer here is that VORP is a stat accumulated across career, and so the longevity of those 3 guys is giving them the edge.

But that avoids the real "smell test" aspect of things, so let me zoom in on career RAPM:

Kawhi 7.5
Giannis 6.6
Sheed 5.7
Divac 5.3
Conley 5.1

I'd expect that most find it hard to accept that Sheed, Divac & Conley are ranking that close to Giannis in career RAPM, and this is specifically where something like a 3-year RAPM peak would be perhaps more likely to give Giannis his due.

But while I have no interesting in trying to say the longevity trio is/was better at basketball than Giannis, I think it is important to recognize that the massive tier-difference that people tend to have between MVP-level guys and guys like Sheed/Divac/Conley isn't as dramatic as people think, and so this is why it's quite common for stars who are say, "misaligned", to see their impact drop below the level of guys who take on less scoring primacy.

And then such players tend to end up with quite resilient value late into their career, which means they get favored more by a state like VORP.

With Kawhi, just briefly, the amount of missed time is just huge at this point when looking at a career accumulation stat like VORP.

[Doctor MJ]

Alright, so I'm going to do a 1996 draft analysis here, as that's the first draft of the PBP era.

I'm going to list guys I've checked based on the -3.3 based VORP. I will also note how this looks differently with the -2.0 based VORP, as well as just RAPM.

1. Ray Allen (pick #5) +3523
2. Kobe Bryant (#13) +3455
3. Steve Nash (#15) +3275
4. Ben Wallace (und) +2151
5. Derek Fisher (#24) +2062
6. Allen Iverson (#1) +1880
7. Peja Stojakovic (#14) +1700
8. Zydrunas Ilgauskas (#20) +987
9. Jermaine O'Neal (#17) +985
10. Marcus Camby (#2) +861

(ftr, other 20K minute guys: Stephon Marbury (AS), Shereef Abdur-Rahim (AS), Antoine Walker (AS), Erick Dampier.)

Top 10 ordering is the same by 2.0 VORP.

By RAPM - as in, not being multiplied by minutes played, we do get some changes:

Nash goes up to #1 with Allen & Kobe each dropping a spot.
Stojakovic moves past Iverson.

If we focus only on Offense:

* For both -3.3 & -2.0 VORP, Kobe takes the #1 spot over Allen & Nash.
* Iverson, Stojakovic, Marbury & Fisher take the next spots, followed by a big gap.
* For RAPM, Nash takes the top spot followed by Allen, with the next 4 in the same order.

If we focus only on Defense, the top 5 by both VORPs & RAPM is the same, and I think it will raise eyebrows:

* Ben tops all 3 lists, and tops by a massive margin (no eyebrows there)
* Fisher takes 2nd on all 3 lists (BIG eyebrows!)
* Jermaine, Camby & Ilgauskas then take spots 3-5.

Okay, so what do we see then?

1. We've got a remarkably close contest between the top 3 of Allen, Kobe & Nash. Nash has the per minute edge on all fronts (overall, offense, defense), but played significantly less than the other two. Kobe having the greatest minutes also allows him to leap Allen for Offensive VORP, but him having the to worst Defensive VORP of the trio allows Alen to squeak in with Overall VORP title.

It probably makes sense at this point to really talk about defense so far as it relates to Kobe & Nash specifically. Kobe had the reputation of a great defender, but rates below RP by this. Nash had the rep as a weak defender, but rates ever so slightly above RP by this. Veterans of these boards will know that this this has been brought up many times in the past. How can Nash possibly come out better than Kobe?

In a nutshell:

a. People need to remember that such data is about what you do day-in and day-out, not about what you can do when called to devote yourself to a new, critical defensive task. While Kobe had the talent for lockdown man defense, that's not generally how he spent his time in the regular season. By contrast, the limited Nash knew he was limited, and played the role assigned to him consistently.

b. People need to remember that "home run" defensive plays can often be more about offensive impact than defensive - at least as so far as metrics like this will award credit. Chief among these is the steal of course, which offensive stars love to gamble for.

A perspective that really hammers this home, of the 6 guys who got 1000 steals in their career from the 1996 draft, 5 rate as below RP on the defensive VORP lists (with both both RP levels). Those 5 are, in the order of career steals: Iverson, Kobe, Allen, Walker & Marbury. The other guy on the steals leaderboard, but with a positive Defensive VORP is Fisher.

We can debate about whether this means we should say these 5 guys were actually "bad defenders", but I think what this definitely says is that they were effectively sacrificing Defensive VORP for Offensive VORP. It might have been the right move overall, but just in terms of giving your the best team defense, it wasn't the right move, which when it seems to speak to not the occasional problematic thief, but basically as a rule to the 1996 class, it really hammers home why it is that shotblockers rather than thieves are considered the gold standard for defensive anchors. (Not that shotblockers can't gamble too much themselves, but they typically aren't looking for a quick bucket when they pressure shots.)

Last note before moving on: Players of the stature of Allen/Kobe/Nash get judged based on the level they reached in their prime more so than career totals certainly, and a look at more specific spans of RAPM is essential when looking to rank these guys, but with this particular data I've put before the group, this is all I'd really want to say. There is more to say of course, but I'm not going to try to at this moment.

2. I don't know if Ben grabbing that 4th spot is necessary to really talk about, but I feel like he just deserves it as arguably the most successful undrafted player of all-time. We know his offense is limited (though by these measures, not below RP) and so in an offense-dominated modern league, it makes sense that the top defensive player is below them, and below them by a good deal.

3. Then we get to Fisher & Iverson, and I think it makes a lot of sense to talk about each of them, and the seemingly insane fact that Fisher comes out on top here. I mean, Iverson played more minutes, so this isn't an obvious case of the inferior player's longevity surpassing a guy with a shortened career.

I think the starting point that can maybe set things at ease is the offense vs defense breakdown. All of this data favors Iverson's offense over Fisher's. So this isn't some situation where we're trying to justify the low primacy role player over the high primacy star as an offensive force. As mentioned before, in addition to Iverson, Stojakovic & Marbury are ahead of Fisher as well - and I think we should note that while Marbury's rep went down the tube later in his career, this data absolutely sees him as an effective offensive star (not an A-lister, but still, having impact with his primacy).

So while this data is literally favoring Fisher's offense over offense-oriented all-stars from his draft class (Walker, Shareef), I might say that it's not saying anything all that radical about Fisher's offense. It's just solid from an impact perspective, not taken by severely cannibalizing team defense, and of course, known for being done on an off-guard role that lets the high primacy star cook. You don't want to draft Fisher to be your offense's fulcrum, and so even after all of this, in a re-draft, picking someone like Iverson/Marbury over Fisher would still make sense.

But then there's the defense. It's Fisher coming out #2 in this draft in defense that's lifting him overall, and that's really something given that he never made All-D while his back court teammate (Kobe) made it as a rule. This data tells such a drastically different story that I really think it deserves a ton of thought from each of us.

If the data's legit, what made Fisher such an effective defender?
If it's skewed somehow, well, how?

4. I'll just tip a cap to Peja at the #6 overall spot & #5 on offense. His career could surely have been greater given his height and shooting, but I'd say he comes off looking here as someone quite solid.

5. We round out the Top 10 with 3 bigs known for defense in Z, Jermaine & Camby. All come off looking certainly good on defense...if a bit underwhelming. As impressive as it is that Fisher ranked 2nd by defense in this cohort for these studies, it's not as impressive as you might think when I say "His defense ranks in front of DPOY Camby!" The reality is that while Camby was a good defender, I don't think he was ever actually a DPOY level guy, and he does stand as an example of a big whose block numbers seem to overstate what he was accomplishing out there.

Z & Jermaine weren't quite seen on that level and so I'm not sure if they are underwhelming here, but all are in a category of quality defensive bigs for the era who shouldn't be confused for the actual outlier defenders of the era (Ben, Duncan, KG).

I will also say about Jermaine: He got supremely overrated at his peak, but this was mostly because of his offense, because he was the #1 option on a 60 win Indiana team, and this helped lead him to become a serious MVP candidate. In reality, this offensive choice (by young Rick Carlisle before seeing the light of modern offensive strategy which, to be fair, was non-existent in the NBA at that time), was questionable at best, and aside from being part of a continued not-quite-enough appreciation for Reggie Miller, it also related to why people were mistakenly seeing Jermaine as the young franchise player for the Pacers, when based on actual talent and impact, it should have been Ron Artest (who then proceeded to take himself out of that conversation with his poor impulse control, which admittedly, didn't come out of nowhere).

6. We now come to the all-stars failing to make the Top 10 (Marbury, Shareef, Walker), who rank 11-13 on the -3.3 VORP. Given the fact these guys, despite playing 28K plus minutes each, fell out of stardom with a whimper, I think this board probably isn't that surprised for their soft landing here.

As I mentioned, Marbury actually does come out in this data like a legit offensive star. While true floor generals like Nash & Jason Kidd would be who he was compared with mostly in his prime, and he would prove lacking to be at their level, that's different from literally being someone running an offense and failing at it. You should want offense better than Marbury's from your best offensive player if you want to contend, but that should be used to infer that most teams actually did have someone better as an offensive alpha. The rub is though, Marbury's Defensive VORP is hideous. It's not just the worse of this draft class, I believe it's literally THE worst Defensive VORP I have on record by either RP level I chose (haven't really tried to "find" the worst though). So then yeah, between the fact that Marbury's offense is quite good but not THAT good, and the fact that his defense was potentially the most damaging of any player of the PBP era, welp, glad you like China Starbury!

In Shareef, we have a candidate for "most overrated by box score" given his splashy entrance as the chosen franchise player of an expansion team (Vancouver) that really, really, really didn't work who then becomes the attempted franchise player of another team (Atlanta) that still didn't work, but got Shareef named all-star. Reef is absolutely a guy we should be keeping in mind when upcoming putative superstars are handed primacy from day 1 - I can't tell you how concerned I am about the Magic betting everything on Paolo. But at the same time, he could look a lot worse by this than he does. No he doesn't look like someone who ever broke through to having actual all-star impact here, but even by the strict -2.0 RP, Shareef still comes off like a better than RP kind of guy.

By contrast Antoine, our trailblazer of the 3-point shot for guys who never learn to shoot, actually has a career RAPM of -2.0, which matches the strict RP exactly making Walker not a below VORP player on our list, but literally a zero. Note that the more generous -3.3 has him as positive, but I think it seems right to talk about Walker as an NBA all-star with a long NBA career who spent enough time actively playing problematic basketball (allowed by his coaches) that he fell at least quite close to career averaging replacement level in its quality. Of course as we do that turning up our nose at his offense, this data says his defense was actually the part that was below RP. Sigh.

7. Last point I'll just point to that last 20K guy, Erick Dampier. Ranked 14th of the 14th 20K guys by -3.3 VORP (though, he is above Walker by -2.2), with below replacement offense, but ranked 6th in Draft for Defensive VORP (Ben, Fisher, Jermaine, Camby, Z). Born in Mississippi, Dampier played 3 years at Miss State, before being drafted 10th overall. He played for 6 teams across 16 seasons, starting in 771 of his 987 games. He probably had his best run in Dallas, which ended when they traded him for Tyson Chandler, and then proceeded to chip the next year (sad trambone). He's listed as having a career earnings of $97 million. He had a good run! Now, his son, Erick Dampier Jr, is currently ranked #3 in the 2028 class by ESPN.

[giberish]

Okay, so what do we see then?

1. We've got a remarkably close contest between the top 3 of Allen, Kobe & Nash. Nash has the per minute edge on all fronts (overall, offense, defense), but played significantly less than the other two. Kobe having the greatest minutes also allows him to leap Allen for Offensive VORP, but him having the to worst Defensive VORP of the trio allows Alen to squeak in with Overall VORP title.

It probably makes sense at this point to really talk about defense so far as it relates to Kobe & Nash specifically. Kobe had the reputation of a great defender, but rates below RP by this. Nash had the rep as a weak defender, but rates ever so slightly above RP by this. Veterans of these boards will know that this this has been brought up many times in the past. How can Nash possibly come out better than Kobe?

In a nutshell:

a. People need to remember that such data is about what you do day-in and day-out, not about what you can do when called to devote yourself to a new, critical defensive task. While Kobe had the talent for lockdown man defense, that's not generally how he spent his time in the regular season. By contrast, the limited Nash knew he was limited, and played the role assigned to him consistently.

I think it's pretty clear that - aside from a few marquee matchups - Kobe's defensive effort during the regular season was limited. This drives down his marks in a regular season study like this. IMO it's also a big reason why the Lakers tended to underachieve somewhat during the regular season and also why they seemed to improve into the playoffs. It wasn't any kind of clutch gene or other BS, it was just a much higher level of defensive effort (Kobe's defensive reputation was based on the playoffs and the occasional marquee regular season game).

As noted, Nash really couldn't coast on defense, he had to put in full effort in order not to be a major problem. Though that meant that he couldn't raise his effort for the playoffs because he was already trying during the regular season.

Though Fisher's defensive numbers here do stand out in a way that I can't explain. I guess the first check would be if they were consistent between his Lakers seasons and his non-Lakers seasons to see if there was something Lakers or Kobe related.

[Doctor MJ]

Okay, so what do we see then?

1. We've got a remarkably close contest between the top 3 of Allen, Kobe & Nash. Nash has the per minute edge on all fronts (overall, offense, defense), but played significantly less than the other two. Kobe having the greatest minutes also allows him to leap Allen for Offensive VORP, but him having the to worst Defensive VORP of the trio allows Alen to squeak in with Overall VORP title.

It probably makes sense at this point to really talk about defense so far as it relates to Kobe & Nash specifically. Kobe had the reputation of a great defender, but rates below RP by this. Nash had the rep as a weak defender, but rates ever so slightly above RP by this. Veterans of these boards will know that this this has been brought up many times in the past. How can Nash possibly come out better than Kobe?

In a nutshell:

a. People need to remember that such data is about what you do day-in and day-out, not about what you can do when called to devote yourself to a new, critical defensive task. While Kobe had the talent for lockdown man defense, that's not generally how he spent his time in the regular season. By contrast, the limited Nash knew he was limited, and played the role assigned to him consistently.

I think it's pretty clear that - aside from a few marquee matchups - Kobe's defensive effort during the regular season was limited. This drives down his marks in a regular season study like this. IMO it's also a big reason why the Lakers tended to underachieve somewhat during the regular season and also why they seemed to improve into the playoffs. It wasn't any kind of clutch gene or other BS, it was just a much higher level of defensive effort (Kobe's defensive reputation was based on the playoffs and the occasional marquee regular season game).

As noted, Nash really couldn't coast on defense, he had to put in full effort in order not to be a major problem. Though that meant that he couldn't raise his effort for the playoffs because he was already trying during the regular season.

Though Fisher's defensive numbers here do stand out in a way that I can't explain. I guess the first check would be if they were consistent between his Lakers seasons and his non-Lakers seasons to see if there was something Lakers or Kobe related.

Good thoughts!

- I totally agree with you that Kobe paced himself in the regular season defensively and so metrics like this don't give us a good understanding of what he was capable of.

- I will reiterate though my point about steals seeming to indicative of poor defense from this perspective. The concept that getting a steal isn't an impactful defensive play is nonsensical, but that doesn't necessarily mean that the play that leads to steal translates to an improved defensive efficiency (DRtg) for the team, and we're seeing it often seems to correlate more improvement to ORtg instead. This is a weird thing that I think everyone needs to understand, and consider what it means when analyzing data such as what I present.

One might decide that wrongly attributing a defense-driven action to offense means that thieves in general are getting their defensive impact underrated by RAPM while their offensive impact is getting overrated by precisely the same amount.

- On Nash, I would actually say he did raise his defensive effort in the playoffs, because I think everybody does. It's certainly possible that Kobe coasted on D in the regular season more than Kobe, but I think we should note that he got about as many steals in the RS as in the PS, so whatever coasting he was doing, it didn't mean he did less gambling, and I think that gambling is probably why his defensive impact numbers never looks that hot.

Again, doesn't mean that gambling was actually hurting the team, but it might have been helping ORtg at the cost of DRtg.

- On Fish, I do want to emphasize that it's not like his career Defensive RAPM is THAT high - it's not like he's anywhere close to Ben Wallace - so it's not like we have to explain why he was secretly a defensive superstar, but I do think his defensive advantage over Kobe by VORP is something large enough that there must be explanations to be had.

- In terms of the correlations in the slices of DRAPM what I see is:

Meh, in the '90s (slight positive)
Absurdly high during the 3-peat. (literally ranks #1 in 2-year DRAPM in 2000 & 2001)
Really falls off in those last couple years before they got rid of Shaq. (actually negative)
Very solid on GS & Utah. (positive)
Falls off a cliff back with the Lakers. (negative)
Once again solid after leaving the Lakers (positive)

I would say that his numbers during the 3-peat absolutely overrate his actual impact and certainly must be inflating his Defensive VORP some, but other than that, his non-Laker years generally look better than his Laker years so we can't really say that the VORP numbers are heavily skewed by his niche on the Lakers.

ftr, trying to pair this up to compare with Kobe doesn't yield any clear correlations. In the 2-year RAPM, Kobe has one sample (2010) where he ranks 57th, beyond that he has no other Top 100 DRAPM samples either in the 2-year range, or 3-year, or 4-year, or 5-year. In fact, there is no 5-year sample where his DRAPM is not negative.

(Note: It occurs to me that I should be clear, "negative" in this case is just meant to represent a "worse than 0" rating from their metric. It shouldn't be taken as a statement of "below replacement" because that was my adjustment, not nbarapm's.)

[Doctor MJ]

Hey y'all,

Not getting a ton of responses here from folks, and I'd really love a discussion, so I'm open to suggestion.

I am however, reluctant to make new threads for every sub-study in the name bringing more people in because it may be that really no one else is that interested in the data, so unless I get a better idea, I'll probably just keep marching through the data, one slice at a time.

Cheers,
Doc

[Doctor MJ]

1997 draft, ranked & rated by generous VORP:

1. Tim Duncan (pick #1) +5259
2. Tracy McGrady (#9) +2179
3. Chauncey Billups (#3) +1752
4. Stephen Jackson (#43) +1069
5. Derek Anderson (#13) +719
6. Antonio Daniels (#4) +669
7. Keith Van Horn (#2) +635
8. Anthony Parker (#21) +633
9. Brevin Knight (#16) +582
10. Tim Thomas (#7) +533

No other all-stars or 20K minute guys

Using the harsh VORP, Parker jumps up from 8 to 6, and Scot Pollard leaps past Thomas.

When we look at RAPM, Pollard gives us the 4th highest number after the big 3 (Duncan, McGrady, Billups) of the draft, followed by Parker.

If we focus only on Offense:

* By both VORPs, the big 3 remain in order, but by RAPM Duncan falls below McGrady & Billups.
* Jackson takes the 4th spot in generous VORP, with Daniels at 5th.
* Daniels takes the 4th spot in harsh VORP & RAPM.

If we focus only on Defense::

* Duncan has a gigantic advantage by all measures, more than doubling anyone else in RAPM, and with far more dramatic advantage by VORPs with his longevity.
* Van Horn, Knight, Pollard & Anderson round out the Top 5 by the VORPs.
* Pollard, Van Horn, Pollard & Anderson round out the Top 5 by RAPM.

Okay, so what do we see then?

1. Tim Duncan is an absolute giant utterly dwarfing all of this draftmates. Shocker I know.

2. Duncan's dominance is predominantly on dominance, but the fact that he has the offensive VORP crown too is informative given that it appears to be based on longevity, but 2010s RAPM slices seem to indicate a much steeper decline in offense than defense, so while Duncan may have an offensive longevity advantage, it's probably not as dramatic as we might theorize.

If I go into nbarapm to look at these slices we've got the option of 2 vs 3 vs 4 vs 5 year, and I definitely have more confidence in the longer spans, but in the name of temporal granularity here, I'm going to focus on the 2-year.

In the 2-year ORAPM sample with the year representing the final year in the span:
Duncan peaks at 5 (2003).
Duncan has 3 spans ranking Top 10 (2003, 2004, 2008).
Duncan has 7 spans ranking Top 30 (earliest 1999, latest 2008).
Duncan has 17 spans ranking Top 100 (earliest 1998, latest 2016). Note that 17 is also the number of years he played.

McGrady peaks at 3 (2003).
McGrady has 3 spans ranking Top 10 (2002, 2003, 2004).
McGrady has 6 spans ranking Top 30 (earliest 2002, latest 2008).
McGrady has 9 years ranking Top 100 (earliest 2000, latest 2008).

Billups peaks at 5 (2009).
Billups has 2 spans ranking Top 10 (2008, 2009)
Billups has 6 spans ranking Top 30 (earliest 2004, latest 2012)
Billups has 10 years ranking Top 100 (earliest 2002, latest 2012)

Okay, well, Duncan's absolutely got a major advantage in offensive longevity by these numbers even if it can't match his own defensive longevity. Wow.

The other two guys by contrast end up with remarkably simple appear longevity despite priming around two very different ages. In 2003, McGrady was 23, while in 2009, Billups (born 3 years earlier) was 32. I'd say it definitely shows that if either a) McGrady had been able to maintain a relatively typical career arc after his precocious start, or b) Billlups had been able to gain traction sooner - like a more typical career arc, that would give one the clear edge over each other, and would probably allow them to top Duncan in Offensive VORP.

- Before I exit McGrady vs Billups, I feel compelled to point out a thing I always try to keep in mind:

A regression stat like RAPM does not make any distinction between making a bad team better or a good team better, and thus doesn't give us a window into whose game scales better in playing with better players, something that we've called "scalability" here for many years. Scalability has a more controversial sister concept we've called "portability" which is harder to define with precision, but speaks to the ability of a player to change how he plays to blend in with other talented teammates, which certainly plays a role in scalability, though the relationship is more complicated that a simple mathematical formula unfortunately.

Anyway, I bring this up to note that Billups strongly proved his scalability in playing starring roles on massively successful playoff teams. By contrast, McGrady's playoff track record for team success is of course, non-exist from a perspective of winning playoff series.

We want to avoid winning bias as a blind thing certainly in our 5-man team sport, but we also know that there are certainly types of things that can lead a player to be more of a floor-raiser (ability to make a bad team better) than a ceiling-raiser (ability to make a good team better), and being a volume scorer with mediocre efficiency is perhaps the #1 thing.

For some perspective on the efficiency gap here, I'll put it like this:

Between the two players, McGrady's 2003 season has the highest TS Add at +192.

Between the two players, they have 9 years with TS Add north of +100. 8 of them were Billups.

Between the two players, they have 18 years with positive TS Add. 14 of them were Billups.

So yeah, huge advantage in shooting efficiency from Billups over their careers, and while that alone won't make you the fulcrum of a serious contender, I don't think Billups would have been able to play that role if were as inefficient as McGrady.

So, just me personally, I would say the VORP metric is underrating Billups relative to McGrady because it's not thinking about scalability-type concerns, and I am, and I see Billups as having the more proven - and accomplished - career even if I'd still have enough faith in '02-03 McGrady to prefer him in a single-year Peak comparison of the type we do in Peaks projects.

- Okay one more thing that I'm to kind of push back against after introducing, but it surprised me in the data to see Billups getting a below RP score for Career VORP. I frankly think it would surprise anyone given that he was part of a legendary defensive core and was named All-D twice. What's up with that?

Well, part of the deal here is Billups' slow start, and part of it has to do with defense falling off hard 2010s, but the really interesting thing is:

a) He never ranks as a Top 100 DRAPM player in any multi-year span I see.

b) Rather than helped by being part of a legendarily defensive 5-man lineup (Biillups/Hamilton/Prince/Wallace/Wallace), his best DRAPM spans come in the back half of the '00s including both the late stage Pistons and his time on the Nuggets afterward.

As we saw in the 1996 draft, Ben Wallace's DRAPM looks great. Rasheed Wallace's data also looks great, and Prince's look quite solid through the Pistons' contending run. Meanwhile the two men of the back court look quite unimpressive (Hamilton looking worse than Billups).

So then this gets us to a question of how an all-time great defensive lineup can have not one, but RP-ish defensive players in 2 of the 5 slots. How can that be right?

Well, my interpretation is that it isn't right, and that the cause of the issue is multicollinearity. For those unfamiliar, the essence of the issue is that if you lack sufficient sample of each player with and without the other, you're going to end up with highly noisy results.

And the thing is, that Pistons 5 played together a ton. Example:

In the '03-04 playoffs, the Piston 5 played 522 minutes together, and no other lineup played more than 50.
In the '24-25 playoffs, the Thunder's primary playoff lineup played 197 minutes, and the team had 3 other lineups with more than 50 minutes.
(Note that both teams went 16-7 in the playoffs.)

That's playoff data rather than RS of course, but believe me it holds true there too. They really had a tendency to just ride that 5-man unit as much as they possibly could.

And these means that the Pistons of this era are the best example I know of, of extreme issue of multicollinearity for teammates in a season span. In fact, these Pistons were specifically on my mind when I reluctantly embraced RAPM over APM as RAPM became the dominant variation, because the regularization process should help with these sort of issues.

And I'm sure it is helping some, but when I look at the 2 to 5 year spans of RAPM during the core Piston contention years, to me that still looks like multicollinearity is making the results something of a funhouse mirror. I think we have plenty of data to conclude Wallace/Wallace/Prince > Billups/Hamilton on defense, but if Billups & Hamilton were actually as ineffective in those lineups as this data appears to say, I honestly don't think you could build a lineup that as epically success as that one was.

So then what am I saying? I believe Billups time in Detroit while it's still a big positive by the VORP stats, he's more likely underrated than overrated by it, and the same is true for Hamilton. By that same token, the other 3 guys might be a bit overrated by it. Though if we're talking specifically about Billups' defense being underrated, I'm sure not looking to jump in and say the Wallaces were overrated defenders, so I'm holding my horses concluding anything too dramatic here.

Alright quickly, after the Big 3 HOFers of the draft, what else do we see?

- Obviously, quite shallow compared to 1996 we know from the start because there's only 5 guys with 20K minutes compared to 14 in 1996.

- And only 4 1000 VORP guys this time with Jackson just barely making it in the generous version, and only 3 making it in the harsh one. Jackson's peak spans appear to largely coincide with his run with Golden State.

- The 4th highest career RAPM of the cohort is Scot Pollard, a career backup. That's weird, right?

So here I'd note that Pollard rates quite highly in his Sacramento run, but unremarkable beyond that. It seems to me that there's not really anything to indicate he should have been massively more utilized (say, graduate to starter minutes) in his time playing for teams other than the Kings.

So what's up with the Sacramento years?

Well, this was the Golden Age of Sacramento basketball with the core that included Webber, Stojakovic, Divac, Bibby, etc, so first and foremost, this was a winning team Pollard was part of, and that generally helps you in all sorts of metrics, but certainly in +/- data.

Now though, as we established with the Pistons, it's possible for RAPM to end up quite harsh when you're one of the lesser lights on that contending constellation, so it's far from a given that Pollard should stand out for the positive at all here, and yet he does. Why?

Well, Pollard's biggest minute year in Sacramento was '01-02. Here were the two primary lineups the team used that year:

Bibby / Christie / Stojakovic / Webber / Divac 770 minutes
Bibby / Christie / Stojakovic / Pollard / Divac 464 minutes

So basically, Webber got injured and Pollard got slotted in in place for him for those 29 games.

The Webber lineup was +4.0 per 100 possessions.
The Pollard lineup was +6.1 per 100 possessions.

Rightly or wrongly, I'd guess that this particular span really helped his career RAPM.

Now, seeing a comparison between lineups like that just in one season, it kinda screams small sample size theater, so point for "wrongly". Proceed with caution.

On the hand, the whole phenomenon of Sacramento seeming to do fine without Webber was kind of a running theme, and while I would say that this meant Webber was indeed overrated, it's not like you could just replace him with a rando and get these results. The Kings in these years had some really effective bigs basically the whole time, and all of those involved with it deserve some respect for that.

- I'll end up by looking at the other guy who stands out just a bit more than the rest by Career RAPM, Anthony "No, I'm not from France" Parker.

Parker was a guy who didn't really get traction in his first 3 years in the NBA, then went off to Europe and became a EuroLeague legend for Maccabi Tel Avis, before coming back to the NBA and playing as a starter from age 31 to 36.

What was different after he came back from Europe? Well, I'm sure he learned a whole lot in his time across the pond(s), but the most salient thing is that he came back an extremely proficient 3-point shooter. And this is where I'd note that this was 2006, and he signed with Toronto who had just hired Bryan Colangelo - formerly of the Suns - as GM. Colangelo's GMing runs after Phoenix have not been successes, but him pulling Parker back from Israel was absolutely a win.

Also of note, Anthony's little sister is women's ball legend Candace Parker, and he is currently the GM of the Orlando Magic.